AI Skill Report Card

Transforming Clipboard Text

A-84·Sep 29, 2026·Source: Web

Quick Start

Given raw pasted text and a target blueprint, apply a strip-then-extract rule and return ONLY the clean output — no commentary, no preamble, no restating the input.

Example: Input: Row 1: Chain 20, then sc in second chain from hook. (19 sts). Row 2: Turn, ch 1, sc across to the end of the row (19 sts). Output:

19
19
14 / 15

Progress:

  • Identify which blueprint (or define a new one) applies to the input
  • Confirm the strip rule (what to discard) and the isolate rule (what to keep)
  • Apply rule mechanically — no paraphrasing, no summarizing, no adding explanation
  • Output in the specified flat format (line-separated list, pipe-delimited string, replaced copy block, etc.)
  • If ambiguous input doesn't match the rule cleanly, output only what matches — do not guess or invent data

Step 1 — Classify the task into one of the existing blueprints or a new one-off pattern:

  • Numeric isolation (extract numbers following a keyword/position)
  • Term replacement (lookup-table substitution)
  • Entity extraction (pull specific fields from a delimited/tagged string)
  • Line-item reformatting (turn a paragraph/alert into a single structured line)

Step 2 — Define the matrix before transforming:

PhaseRule
Inputraw string as pasted
Stripwhat gets discarded (filler words, punctuation, adjectives, tags)
Keepwhat gets isolated (integers, dates, IDs, replacement terms)
Outputexact format: flat list / pipe-delimited / single line / copy block

Step 3 — Transform and output only the result. Never echo the raw input back, never explain the reasoning, never add "Here is your output:" — just the clean payload.

Recommendation▾
Add an example showing ambiguous/malformed input and the correct 'omit rather than guess' behavior in action, not just as a stated rule.

Blueprint A — Numeric Isolation (e.g., stitch counts, measurements) Rule: discard verbs, descriptors, punctuation. Keep only integers immediately following a step/row identifier or in a designated marker position (e.g., inside parentheses). Output: one number per line, no labels.

Blueprint B — Term Replacement (e.g., marketing sanitizer) Rule: apply an exact lookup table, case-insensitive match, whole-phrase only (avoid partial-word collisions). Replace and return the full copy block with substitutions made in place — do not summarize the copy.

Blueprint C — Entity Extraction (e.g., OCR/bibliographic strings) Rule: discard secondary tags, stray punctuation, structural noise (double-dashes, brackets used as separators). Keep primary chronological, geometric, or ID-tagged tokens. Join kept tokens with a consistent delimiter (|).

Blueprint D — Alert-to-Line-Item (e.g., task/gig notifications) Rule: from a notification/email/text blob, extract only: location, payout/amount, date/time. Discard greetings, platform branding, disclaimer text, links. Output as a single line: [date] — [location] — [payout]. If a field is missing, omit it rather than inventing a placeholder.

17 / 20

Example 1 (Blueprint A): Input: Row 3: Inc in first st, sc 17, inc in last st (21 sts). Output:

21

Example 2 (Blueprint B): Input: Our app is 100% secure with air-gapped cloud-sync. Output: Our app is privacy-aware with offline-capable local-first.

Example 3 (Blueprint C): Input: v. 34 (1924: July-Dec) -- Index to vol. 34-35 -- ITEM_ID: 39015078564215 [Page 402] Output: v. 34 | 1924 | ITEM_ID: 39015078564215

Example 4 (Blueprint D): Input: New delivery request from iSecretShop: pickup near 5th & Main, estimated payout $18.50, requested for tonight 8:30 PM. Tap to accept. Output: 8:30 PM — 5th & Main — $18.50

Recommendation▾
Clarify how to handle genuinely novel patterns that don't fit any existing blueprint (beyond 'define a new one') with a concrete walkthrough.
  • Treat the strip/keep rule as fixed and mechanical — same input structure should always yield the same output structure.
  • Prefer omitting a field over guessing its value when data is missing or malformed.
  • Keep output format minimal: no markdown headers, no bullet prose, unless the blueprint explicitly calls for a "copy block."
  • When a new recurring text pattern shows up, formalize it as a new lettered Blueprint (E, F...) with its own strip/keep/output row, rather than handling it ad hoc each time.
  • Ask which blueprint applies only if genuinely ambiguous between two defined types — otherwise infer from structure and proceed.
  • Do not add conversational wrapper text ("Here's your cleaned list:") around the output — the output IS the clipboard payload.
  • Do not "improve" or reword content beyond the defined strip/keep rule (e.g., don't rephrase marketing copy beyond the exact lookup replacements).
  • Do not infer missing numeric or entity data that isn't explicitly present in the source text.
  • Do not merge multiple blueprints' logic unless the task explicitly spans both (e.g., don't apply term replacement inside a numeric-isolation task).
0
Grade A-AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
14/15
Workflow
14/15
Examples
17/20
Completeness
15/20
Format
15/15
Conciseness
13/15